Validate before launch

Test Sagent's catalog, conversations, and entry points before publishing.

Validation confirms that Sagent works as intended before shoppers use it. It catches catalog, conversation, and deploy issues while they are still safe to fix.

Use this checklist to test the journeys your shoppers are most likely to take. Resolve unexpected results before publishing a version.

1

Check catalog coverage

In Products, use search and filters to verify products appear for different property combinations. Repeat the check for every supported locale.

Verify each result has the correct title, price, image, availability, and product link.

2

Test conversations

Use preview mode in Setup to test real queries. Test simple queries, ambiguous requests, and complex requests.

Test both flows:

• Start Product Flow with a product already in context.

• Start Guided Flow with the full catalog in scope.

Test configured capabilities, including Comparison and Review summarizer.

3

Run the pre-release test matrix

Create one row for each journey before testing. Record the expected outcome, actual outcome, tester, device, and locale.

To check	Test	Expected outcome
Catalog and locale	Search key property combinations in each locale.	Only eligible products appear with localized data.
Guided Flow	Test simple, ambiguous, and complex requests.	The flow asks relevant questions and recommends suitable products.
Product Flow	Start with a product in context.	The flow uses that product context correctly.
Enabled tools	Test each enabled tool, such as comparison or review summaries.	Each tool opens and returns relevant content.
Product links and CTAs	Open every product link and CTA.	Links reach the intended destination and CTAs work.
Disclosure	Start every deploy type and entry point.	The AI disclosure appears before the first message.
Responsive design	Repeat priority journeys on mobile and desktop.	Content remains readable and actions remain usable.

Test every live entry point and deploy type. Include page types, markets, and languages that use Sagent.

4

Investigate failures

Do not publish while a critical journey fails. Link each failed matrix row to its corrective area:

• Products: Missing, incorrect, unavailable, or poorly localized product data.

• Knowledge or guidance: Incorrect answers, weak questions, or unsuitable recommendations.

• Behavior or design: Broken tools, links, CTAs, disclosure, entry points, or responsive layouts.

• Deploy: Incorrect targeting, locale, version, or publishing configuration.

Retest the exact row after the correction. Publish only when every required row passes.

5

Check disclosure and privacy messaging

Open every deploy type. Confirm the AI disclosure appears before the first user message. Confirm privacy content and links are visible and correct.

6

Monitor after launch

Review Monitoring after major updates. Investigate flagged or unexpected conversations and refine guidance or product data.

Frequently asked questions

Why should I test simple and complex queries?

Testing catches data gaps and question-flow issues before shoppers encounter them. Complex queries also reveal unclear guidance and unexpected responses.

What should I check at every entry point?

Check that Sagent opens correctly, the AI disclosure is visible, the intended flow starts, and recommendations use the expected catalog.

Why validation matters

Sagent can misinterpret answers or surface unsuitable products when data or guidance has gaps. Test simple and complex scenarios to find those issues before launch.

What's next

• Monitoring

• Analytics

• Setting up a Product Feed